Maciej Kulesza

Gaze-tracking-based audio-visual correlation analysis employing quality of experience methodology

This paper investigates a new approach to audio-visual correlation assessment based on the gaze-tracking system developed at the Multimedia Systems Department (MSD) of Gdansk University of Technology (GUT). The gaze-tracking methodology, having roots in Human-Computer Interaction borrows the relevance feedback through gaze-tracking and applies it to the new area of interests, which is Quality of Experience (QoE). Results of subjective tests carried out at the MSD showed a strong dependency between video presented in the screen and the perceived audio. It has also been shown that the application of gaze-tracking to the audio-visual correlation analysis allows for the objectivization of results obtained in subjective tests. Therefore this research study concentrates on the possibility to apply this methodology to the area of Quality of Experience.

Fast communication: A low complexity double-talk detector based on the signal envelope

A new algorithm for double-talk detection, intended for use in the acoustic echo canceller for voice communication applications, is proposed. The communication system developed by the authors required the use of a double-talk detection algorithm with low complexity and good accuracy. The authors propose an approach to double-talk detection based on the signal envelopes. For each of three signals: the far-end speech, the microphone signal and the echo estimate, an envelope is detected. Next, using these envelopes, a detection function is determined and compared to the threshold. Additionally, a dynamic threshold is introduced in order to improve the accuracy of the algorithm. The results of the simulations presented in the paper proved that the accuracy of double-talk detection obtained using the proposed algorithm is higher than in the Geigel algorithm and comparable to the correlation-based methods, while the computational complexity of the proposed method remains at an acceptable level. The double-talk detection algorithm presented here may be used in voice communication systems having limited resources, allowing for accurate double-talk detection and, as a consequence, efficient acoustic echo cancellation.

A novel tonality estimation method of spectral components for audio coding applications.

A novel algorithm providing adequate tonality estimates for constant and modulated sinusoidal components was developed. The proposed algorithm was combined with the MPEG psychoacoustic model 2 in order to verify whether replacing the standard tonality estimator with the proposed one leads to a more reliable estimate of hearing threshold. It was verified whether the proposed tonality estimation algorithm may be used as a basis for detecting the signal bands containing only a noise‐like component that can be encoded according to the perceptual noise substitution technique. Subsequently, it was proved that it is possible to estimate the tonality of unmodulated or frequency‐modulated sinusoidal components of audio signals through the comparison of their instantaneous frequency variations determined employing both an estimator processing spectral amplitude samples and an estimator processing spectral phase samples. It was also revealed that distortions introduced during perceptual audio coding may be effective...

Improving signal quality of a speech codec using hybrid perceptual-parametric algorithm

A hybrid parametric-perceptual speech codec architecture is presented. The basic CELP parametric codec structure is enhanced using the perceptual coding method. The objective of the codec hybridisation is obtaining significant improvement in the perceived signal quality. Two hybrid architectures are proposed. The first one encodes voiced parts perceptually in the CELP residual signal. The second one divides the signal into voiced signal components that are encoded using the perceptual algorithm, unvoiced components that are encoded parametrically and transients remaining unencoded. The signal quality achieved using the hybrid codec is compared to the quality of some standard speech codecs.

Transient detection algorithms for speech coding applications

One of the methods for improving transmitted speech quality is efficient encoding of transients. The majority of currently used speech codecs employ the parametric approach, thus they do not differentiate between transients and other signal components. A sine+noise+transients signal model, proposed earlier as a tool for wideband audio signal encoding, seems to be a promising solution for enhancing signal quality in speech coding. Methods of transient identifying for an efficient transform‐domain encoding of transients has to be chosen carefully. Different transient detection algorithms are described in literature, most of them optimized for musical signals. The methods do not ensure efficient transient detection in speech signal, due to the different character of musical signals and speech. This paper provides a comparison of several transient detection methods, mostly based on evaluation of signal energy, that are now primarily used for musical signals. Their efficiency and application to transient detec...